Derivative-Free Optimization Methods

A flexible and expandable library for derivative-free optimization algorithms.

Carlos Cotta, 2024

Features

Currently included methods are:

• Hooke-Jeeves algorithm
• Nelder-Mead algorithm

Requirements

Requires JDK 17 or higher. The project relies on the json-simple library for parsing configuration files and dumping statistics. A Maven dependency is included in the pom.xml file for this purpose. Alternatively, non-Maven users may download json-simple-4.0.1.jar and add it to the project build path.

Usage

See dfopt/test/RunDerivativeFree class for an example of use.

The configuration of the algorithm is done via a JSON file (see run/hookejeeves.json and run/neldermead.json for configuration examples for each of the algorithms).

If you are using Maven, the following dependency can be added to your project:

    <dependency>
    	<groupId>es.uma.lcc.caesium</groupId>
    	<artifactId>derivative-free-optimization</artifactId>
  	<version>1.0</version>
    </dependency>

Applications

The library has been used in the following projects (non-exhaustive list):

• pedestrian-evacuation-optimization: Optimization of crowd evacuation in case of emergency